9 docs tagged with "nebula"

Testing vLLM and SGLang against Ollama on Tanker (Maxwell) and Galleon (Ampere) — and why matching the compute capability isn't enough.

How a handful of libvirt XML tweaks turned a sluggish Windows Server 2025 VM into a high-performance guest — with benchmark numbers to prove it.

Every Matey spawn starts from zero — no memory of the last task, no memory of the one before. Trimming what it's told to remember, on every single call, turned out to matter more than which model it runs on.

Two fleet-facing services lived as bare-metal processes with no lifecycle control. A Podman Containerfile, a pair of Quadlet units, and two false "no-go" alarms later, both run containerized with zero loss of function.

How a 240Hz DisplayPort link drop became a keyboard-breaking DPMS rabbit hole — and how we built a two-key fix that survived a reboot.

Where standalone ideas form before they own a galaxy.

Benchmarking 12B models across the fleet — MLX, CPU, and Vulkan — and the ship role reassignment that followed.

Goose, agy, and a Python script walk into the same terminal. Only one leaves at a useful speed.

How we tamed gemma4:12b-mlx verbosity from 1537 tokens to 59 — and what it revealed about system prompt engineering.